Added custom egress for model deployment

Author: lu-ohai

ads/model/deployment/model_deployment.py

@@ -1520,6 +1520,9 @@ def _build_model_deployment_configuration_details(self) -> Dict:
                 or MODEL_DEPLOYMENT_INSTANCE_MEMORY_IN_GBS,
             }
 
+        if infrastructure.subnet_id:
+            instance_configuration[infrastructure.CONST_SUBNET_ID] = infrastructure.subnet_id
+
         scaling_policy = {
             infrastructure.CONST_POLICY_TYPE: "FIXED_SIZE",
             infrastructure.CONST_INSTANCE_COUNT: infrastructure.replica

ads/model/deployment/model_deployment_infrastructure.py

@@ -39,6 +39,8 @@ class ModelDeploymentInfrastructure(Builder):
         The access and predict log id of model deployment
     web_concurrency: int
         The web concurrency of model deployment
+    subnet_id: str
+        The subnet id of model deployment
 
     Methods
     -------
@@ -64,6 +66,8 @@ class ModelDeploymentInfrastructure(Builder):
         Sets the access and predict log id of model deployment
     with_web_concurrency(web_concurrency)
         Sets the web concurrency of model deployment
+    with_subnet_id(subnet_id) 
+        Sets the subnet id of model deployment
 
     Example
     -------
@@ -79,6 +83,7 @@ class ModelDeploymentInfrastructure(Builder):
     ...        .with_replica(1)
     ...        .with_bandwidth_mbps(10)
     ...        .with_web_concurrency(10)
+    ...        .with_subnet_id(<subnet_id>)
     ...        .with_access_log(
     ...            log_group_id=<log_group_id>,
     ...            log_id=<log_id>
@@ -121,6 +126,7 @@ class ModelDeploymentInfrastructure(Builder):
     CONST_LOG_GROUP_ID = "logGroupId"
     CONST_WEB_CONCURRENCY = "webConcurrency"
     CONST_STREAM_CONFIG_DETAILS = "streamConfigurationDetails"
+    CONST_SUBNET_ID = "subnetId"
 
     attribute_map = {
         CONST_PROJECT_ID: "project_id",
@@ -136,6 +142,7 @@ class ModelDeploymentInfrastructure(Builder):
         CONST_LOG_ID: "log_id",
         CONST_LOG_GROUP_ID: "log_group_id",
         CONST_WEB_CONCURRENCY: "web_concurrency",
+        CONST_SUBNET_ID: "subnet_id"
     }
 
     shape_config_details_attribute_map = {
@@ -162,6 +169,7 @@ class ModelDeploymentInfrastructure(Builder):
         CONST_COMPARTMENT_ID: "compartment_id",
         CONST_SHAPE_NAME: f"{MODEL_CONFIG_DETAILS_PATH}.instance_configuration.instance_shape_name",
         CONST_SHAPE_CONFIG_DETAILS: f"{MODEL_CONFIG_DETAILS_PATH}.instance_configuration.model_deployment_instance_shape_config_details",
+        CONST_SUBNET_ID: f"{MODEL_CONFIG_DETAILS_PATH}.instance_configuration.subnet_id",
         CONST_REPLICA: f"{MODEL_CONFIG_DETAILS_PATH}.scaling_policy.instance_count",
         CONST_BANDWIDTH_MBPS: f"{MODEL_CONFIG_DETAILS_PATH}.bandwidth_mbps",
         CONST_ACCESS_LOG: "category_log_details.access",
@@ -518,3 +526,29 @@ def with_web_concurrency(
             The ModelDeploymentInfrastructure instance (self).
         """
         return self.set_spec(self.CONST_WEB_CONCURRENCY, web_concurrency)
+
+    def with_subnet_id(self, subnet_id: str) -> "ModelDeploymentInfrastructure":
+        """Sets the subnet id of model deployment.
+
+        Parameters
+        ----------
+        subnet_id : str
+            The subnet id of model deployment.
+
+        Returns
+        -------
+        ModelDeploymentInfrastructure
+            The ModelDeploymentInfrastructure instance (self).
+        """
+        return self.set_spec(self.CONST_SUBNET_ID, subnet_id)
+
+    @property
+    def subnet_id(self) -> str:
+        """The model deployment subnet id.
+        
+        Returns
+        -------
+        str
+            The model deployment subnet id.
+        """
+        return self.get_spec(self.CONST_SUBNET_ID, None)

ads/model/generic_model.py

@@ -1902,6 +1902,7 @@ def deploy(
         display_name: Optional[str] = None,
         description: Optional[str] = None,
         deployment_instance_shape: Optional[str] = None,
+        deployment_instance_subnet_id: Optional[str] = None,
         deployment_instance_count: Optional[int] = None,
         deployment_bandwidth_mbps: Optional[int] = None,
         deployment_log_group_id: Optional[str] = None,
@@ -1950,6 +1951,8 @@ def deploy(
             The description of the model.
         deployment_instance_shape: (str, optional). Default to `VM.Standard2.1`.
             The shape of the instance used for deployment.
+        deployment_instance_subnet_id: (str, optional). Default to None.
+            The subnet id of the instance used for deployment.
         deployment_instance_count: (int, optional). Defaults to 1.
             The number of instance used for deployment.
         deployment_bandwidth_mbps: (int, optional). Defaults to 10.
@@ -2084,6 +2087,10 @@ def deploy(
                 self.properties.deployment_instance_shape
                 or existing_infrastructure.shape_name
             )
+            .with_subnet_id(
+                self.properties.deployment_instance_subnet_id
+                or existing_infrastructure.subnet_id
+            )
             .with_replica(
                 self.properties.deployment_instance_count
                 or existing_infrastructure.replica
@@ -2253,6 +2260,7 @@ def prepare_save_deploy(
         deployment_display_name: Optional[str] = None,
         deployment_description: Optional[str] = None,
         deployment_instance_shape: Optional[str] = None,
+        deployment_instance_subnet_id: Optional[str] = None,
         deployment_instance_count: Optional[int] = None,
         deployment_bandwidth_mbps: Optional[int] = None,
         deployment_log_group_id: Optional[str] = None,
@@ -2340,6 +2348,8 @@ def prepare_save_deploy(
             The description of the model.
         deployment_instance_shape: (str, optional). Default to `VM.Standard2.1`.
             The shape of the instance used for deployment.
+        deployment_instance_subnet_id: (str, optional). Default to None.
+            The subnet id of the instance used for deployment.
         deployment_instance_count: (int, optional). Defaults to 1.
             The number of instance used for deployment.
         deployment_bandwidth_mbps: (int, optional). Defaults to 10.
@@ -2481,6 +2491,7 @@ def prepare_save_deploy(
             display_name=deployment_display_name,
             description=deployment_description,
             deployment_instance_shape=self.properties.deployment_instance_shape,
+            deployment_instance_subnet_id=self.properties.deployment_instance_subnet_id,
             deployment_instance_count=self.properties.deployment_instance_count,
             deployment_bandwidth_mbps=self.properties.deployment_bandwidth_mbps,
             deployment_log_group_id=self.properties.deployment_log_group_id,

ads/model/model_properties.py

@@ -28,6 +28,7 @@ class ModelProperties(BaseProperties):
     remove_existing_artifact: bool = None
     overwrite_existing_artifact: bool = None
     deployment_instance_shape: str = None
+    deployment_instance_subnet_id: str = None
     deployment_instance_count: int = None
     deployment_bandwidth_mbps: int = None
     deployment_log_group_id: str = None

tests/unitary/default_setup/model_deployment/test_model_deployment_v2.py

@@ -79,6 +79,7 @@
                 model_deployment_instance_shape_config_details=ModelDeploymentInstanceShapeConfigDetails(
                     ocpus=10, memory_in_gbs=36
                 ),
+                subnet_id="fakeid.subnet.oc1.iad.xxx"
             ),
             scaling_policy=FixedSizeScalingPolicy(instance_count=5),
             bandwidth_mbps=5,
@@ -128,6 +129,7 @@
                 "replica": 5,
                 "shape_name": "VM.Standard.E4.Flex",
                 "shape_config_details": {"ocpus": 10, "memoryInGBs": 36},
+                "subnet_id": "fakeid.subnet.oc1.iad.xxx",
                 "web_concurrency": 10,
                 "access_log": {
                     "logGroupId": "fakeid.loggroup.oc1.iad.xxx",
@@ -184,6 +186,7 @@
       shapeConfigDetails:
         memoryInGBs: 36
         ocpus: 10
+      subnetId: fakeid.subnet.oc1.iad.xxx
       replica: 5
       bandwidthMbps: 5
       webConcurrency: 5
@@ -215,6 +218,7 @@ def initialize_model_deployment(self):
             .with_replica(5)
             .with_shape_name("VM.Standard.E4.Flex")
             .with_shape_config_details(ocpus=10, memory_in_gbs=36)
+            .with_subnet_id("fakeid.subnet.oc1.iad.xxx")
             .with_web_concurrency(10)
             .with_access_log(
                 log_group_id="fakeid.loggroup.oc1.iad.xxx",
@@ -267,6 +271,7 @@ def initialize_model_deployment_from_spec(self):
             .with_replica(5)
             .with_shape_name("VM.Standard.E4.Flex")
             .with_shape_config_details(ocpus=10, memory_in_gbs=36)
+            .with_subnet_id("fakeid.subnet.oc1.iad.xxx")
             .with_web_concurrency(10)
             .with_access_log(
                 log_group_id="fakeid.loggroup.oc1.iad.xxx",
@@ -318,6 +323,7 @@ def initialize_model_deployment_from_kwargs(self):
             .with_replica(5)
             .with_shape_name("VM.Standard.E4.Flex")
             .with_shape_config_details(ocpus=10, memory_in_gbs=36)
+            .with_subnet_id("fakeid.subnet.oc1.iad.xxx")
             .with_web_concurrency(10)
             .with_access_log(
                 log_group_id="fakeid.loggroup.oc1.iad.xxx",
@@ -404,6 +410,7 @@ def test_initialize_model_deployment(self):
             "ocpus": 10,
             "memoryInGBs": 36,
         }
+        assert temp_infrastructure.subnet_id == "fakeid.subnet.oc1.iad.xxx"
         assert temp_infrastructure.replica == 5
         assert temp_infrastructure.access_log == {
             "logGroupId": "fakeid.loggroup.oc1.iad.xxx",
@@ -461,6 +468,7 @@ def test_model_deployment_to_dict(self):
                         "replica": 5,
                         "shapeName": "VM.Standard.E4.Flex",
                         "shapeConfigDetails": {"ocpus": 10, "memoryInGBs": 36},
+                        "subnetId": "fakeid.subnet.oc1.iad.xxx",
                         "webConcurrency": 10,
                         "accessLog": {
                             "logGroupId": "fakeid.loggroup.oc1.iad.xxx",
@@ -517,6 +525,7 @@ def test_build_model_deployment_configuration_details(self, mock_prepare_artifac
                         "ocpus": 10,
                         "memoryInGBs": 36,
                     },
+                    "subnetId": "fakeid.subnet.oc1.iad.xxx",
                 },
                 "modelId": "fakeid.datasciencemodel.oc1.iad.xxx",
                 "scalingPolicy": {"policyType": "FIXED_SIZE", "instanceCount": 5},
@@ -680,6 +689,7 @@ def test_build_model_deployment_details(self, mock_prepare_artifact):
             instance_configuration.model_deployment_instance_shape_config_details.memory_in_gbs
             == model_deployment.infrastructure.shape_config_details["memoryInGBs"]
         )
+        assert instance_configuration.subnet_id == model_deployment.infrastructure.subnet_id
 
         scaling_policy = model_configuration_details.scaling_policy
         assert isinstance(scaling_policy, FixedSizeScalingPolicy)
@@ -780,6 +790,7 @@ def test_update_from_oci_model(self):
             infrastructure.shape_config_details["memoryInGBs"]
             == instance_configuration.model_deployment_instance_shape_config_details.memory_in_gbs
         )
+        assert infrastructure.subnet_id == instance_configuration.subnet_id
         assert infrastructure.replica == scaling_policy.instance_count
 
         category_log_details = OCI_MODEL_DEPLOYMENT_RESPONSE.category_log_details
@@ -860,6 +871,7 @@ def test_model_deployment_from_yaml(self):
                         "replica": 5,
                         "shapeName": "VM.Standard.E4.Flex",
                         "shapeConfigDetails": {"ocpus": 10, "memoryInGBs": 36},
+                        "subnetId": "fakeid.subnet.oc1.iad.xxx",
                         "accessLog": {
                             "logGroupId": "fakeid.loggroup.oc1.iad.xxx",
                             "logId": "fakeid.log.oc1.iad.xxx",
@@ -1026,6 +1038,7 @@ def test_update_model_deployment_details(self, mock_prepare_artifact):
             instance_configuration.model_deployment_instance_shape_config_details.memory_in_gbs
             == model_deployment.infrastructure.shape_config_details["memoryInGBs"]
         )
+        assert instance_configuration.subnet_id == model_deployment.infrastructure.subnet_id
 
         scaling_policy = model_configuration_details.scaling_policy
         assert isinstance(scaling_policy, FixedSizeScalingPolicy)
@@ -1077,6 +1090,7 @@ def test_extract_from_oci_model(self):
                 "replica": 5,
                 "shapeName": "VM.Standard.E4.Flex",
                 "shapeConfigDetails": {"ocpus": 10, "memoryInGBs": 36},
+                "subnetId": "fakeid.subnet.oc1.iad.xxx",
                 "accessLog": {
                     "logGroupId": "fakeid.loggroup.oc1.iad.xxx",
                     "logId": "fakeid.log.oc1.iad.xxx",

tests/unitary/with_extras/model/test_generic_model.py

@@ -520,6 +520,7 @@ def test_deploy_success(self, mock_deploy):
         infrastructure = ModelDeploymentInfrastructure(
             **{
                 "shape_name": "test_deployment_instance_shape",
+                "subnet_id": "test_deployment_subnet_id",
                 "replica": 10,
                 "bandwidth_mbps": 100,
                 "shape_config_details": {"memory_in_gbs": 10, "ocpus": 1},
@@ -560,6 +561,7 @@ def test_deploy_success(self, mock_deploy):
             "display_name": "test_display_name",
             "description": "test_description",
             "deployment_instance_shape": "test_deployment_instance_shape",
+            "deployment_instance_subnet_id": "test_deployment_subnet_id",
             "deployment_instance_count": 10,
             "deployment_bandwidth_mbps": 100,
             "deployment_memory_in_gbs": 10,
@@ -604,6 +606,7 @@ def test_deploy_success(self, mock_deploy):
             "ocpus": input_dict["deployment_ocpus"],
             "memory_in_gbs": input_dict["deployment_memory_in_gbs"],
         }
+        assert result.infrastructure.subnet_id == input_dict["deployment_instance_subnet_id"]
         assert result.runtime.image == input_dict["deployment_image"]
         assert result.runtime.entrypoint == input_dict["entrypoint"]
         assert result.runtime.server_port == input_dict["server_port"]
@@ -1328,6 +1331,7 @@ def test__to_yaml(self):
                     "display_name": "fake_deployment_display_name",
                     "description": None,
                     "deployment_instance_shape": None,
+                    "deployment_instance_subnet_id": None,
                     "deployment_instance_count": None,
                     "deployment_bandwidth_mbps": None,
                     "deployment_log_group_id": None,
@@ -1367,6 +1371,7 @@ def test__to_yaml(self):
                     "deployment_display_name": "fake_deployment_display_name",
                     "deployment_description": None,
                     "deployment_instance_shape": None,
+                    "deployment_instance_subnet_id": None,
                     "deployment_instance_count": None,
                     "deployment_bandwidth_mbps": None,
                     "deployment_log_group_id": None,
@@ -1417,6 +1422,7 @@ def test__to_yaml(self):
                     "display_name": "fake_deployment_display_name",
                     "description": None,
                     "deployment_instance_shape": None,
+                    "deployment_instance_subnet_id": None,
                     "deployment_instance_count": None,
                     "deployment_bandwidth_mbps": None,
                     "deployment_log_group_id": None,
@@ -1456,6 +1462,7 @@ def test__to_yaml(self):
                     "deployment_display_name": "fake_deployment_display_name",
                     "deployment_description": "fake_deployment_description",
                     "deployment_instance_shape": "2.1",
+                    "deployment_instance_subnet_id": "ocid1.subnet.oc1.iad.<unique_ocid>",
                     "deployment_instance_count": 1,
                     "deployment_bandwidth_mbps": 10,
                     "deployment_log_group_id": "ocid1.loggroup.oc1.iad.<unique_ocid>",
@@ -1512,6 +1519,7 @@ def test__to_yaml(self):
                     "deployment_instance_shape": "2.1",
                     "deployment_instance_count": 1,
                     "deployment_bandwidth_mbps": 10,
+                    "deployment_instance_subnet_id": "ocid1.subnet.oc1.iad.<unique_ocid>",
                     "deployment_log_group_id": "ocid1.loggroup.oc1.iad.<unique_ocid>",
                     "deployment_access_log_id": "ocid1.log.oc1.iad.<unique_ocid>",
                     "deployment_predict_log_id": "ocid1.log.oc1.iad.<unique_ocid>",
@@ -1553,6 +1561,7 @@ def test__to_yaml(self):
                     "deployment_display_name": "fake_deployment_display_name",
                     "deployment_description": "fake_deployment_description",
                     "deployment_instance_shape": "2.1",
+                    "deployment_instance_subnet_id": "ocid1.subnet.oc1.iad.<unique_ocid>",
                     "deployment_instance_count": 1,
                     "deployment_bandwidth_mbps": 10,
                     "deployment_log_group_id": "ocid",
@@ -1615,6 +1624,7 @@ def test__to_yaml(self):
                     "deployment_instance_shape": "2.1",
                     "deployment_instance_count": 1,
                     "deployment_bandwidth_mbps": 10,
+                    "deployment_instance_subnet_id": "ocid1.subnet.oc1.iad.<unique_ocid>",
                     "deployment_log_group_id": "ocid",
                     "deployment_access_log_id": "ocid",
                     "deployment_predict_log_id": "ocid",
@@ -1685,6 +1695,7 @@ def test_prepare_save_deploy_with_default_display_name(
             "display_name": utils.get_random_name_for_resource(),
             "description": None,
             "deployment_instance_shape": None,
+            "deployment_instance_subnet_id": None,
             "deployment_instance_count": None,
             "deployment_bandwidth_mbps": None,
             "deployment_memory_in_gbs": None,